2026-06-30 20:02:09
Welcome to the Premium edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
The May 2026 IPO priced at $185, surged to $386 on opening day, and closed at $311, a 68% pop.
That was the high-water mark. The stock spent the next six weeks giving back most of its gains, and its first earnings report triggered a record two-day plunge that briefly pushed it below the IPO price.
If that round-trip feels familiar, it’s the pattern we walked through in How To Invest in IPOs. The day-one pop is often short-lived, and the first earnings call is where the hype meets reality.
The quarter itself was impressive. Revenue nearly doubled, Cerebras beat expectations, and management raised full-year guidance above consensus.
The problem was not demand. It was the cost of keeping up with it.
Management argues the market misunderstood the softer margin outlook, which is partly due to Cerebras renting back its own chips from a customer while it waits for new data centers to come online.
Let’s review what we learned.
Today at a glance:
🧠 How Cerebras makes money
📊 Q1 FY26 by the numbers
🎙️ Key insights from the call
🔭 What to watch next
FROM OUR PARTNERS
When McKinsey & Company projects $7.1 trillion in AI investment by 2030, it’s not meant just for names like OpenAI, Anthropic, and Google.
Investors want to know the new movers in this industry. That’s why they’re looking at companies like BluSky AI.
These companies are building the infrastructure layer of the AI future. And their tech is only getting more important with time as demand for AI compute power outruns supply.
While typical data centers take several years to build, BluSky AI deploys modular, AI-ready data centers in months. BluSky AI utilizes less power, little to no water, and minimizes the impact on local communities.
Startups, enterprises, and governments will depend on tech like this. BluSky AI has a robust portfolio of unique sites with negotiated power ready for deployment. Become an early-stage investor in BlueSky AI today.
This is a paid advertisement for BluSky AI Regulation A offering. Please read the offering circular at https://invest.bluskyaidatacenters.com/
We’ve not covered Cerebras before in this newsletter, so let’s go back to basics.
A typical AI processor is roughly the size of a postage stamp. Cerebras builds one the size of a dinner plate: an entire silicon wafer turned into a single chip. Its Wafer-Scale Engine packs ~4 trillion transistors, against ~200 billion on NVIDIA's flagship Blackwell. Cerebras sells it inside a single machine, the CS-3, rather than as a bare chip.
The point of all that silicon is speed. Keeping a model on one giant chip avoids the slow shuffle of data between thousands of smaller ones, which lets Cerebras generate tokens roughly an order of magnitude faster than a GPU cluster. On frontier models, it clears 1,000 tokens per second. The pitch is simple: fast AI is worth a premium because it gets more done.
Cerebras turns that into two revenue streams.
Hardware: Ceberas sells CS-3 systems and clusters to customers who run them in their own data centers. This segment was $111 million in Q1 FY26 (+59% Y/Y), and it’s the legacy side of the business.
Cloud & other services: Customers can rent that same inference horsepower by the token from Cerebras-operated data centers. This segment hit $83 million (+178% Y/Y), growing three times faster than hardware and the strategic center of gravity going forward.
The moat is where it gets interesting. Cerebras sidesteps the exact chokepoints throttling everyone else. It uses no high-bandwidth memory (HBM), no CoWoS advanced packaging, and no bleeding-edge 3nm process, leaning instead on plentiful SRAM and mature 5nm. While the industry fights over the same scarce TSMC and HBM allocation, Cerebras isn’t standing in that line, and it’s the only AI accelerator maker building its systems entirely in the US.
The catch is who’s been paying. Cerebras was, until recently, almost entirely dependent on G42, an Abu Dhabi AI group. Its risk disclosures now name four significant customers: OpenAI, G42, MBZUAI University, and AWS. The bull case rests on the first and last names on that list, turning a UAE-heavy customer base into broader Western enterprise demand.
One quirk of the OpenAI deal shapes the numbers throughout, so it’s worth understanding up front:
The deal: A multi-year agreement for more than $20 billion of Cerebras compute.
The twist: To win it, Cerebras also handed OpenAI warrants (the right to buy its stock at a discount).
The accounting: Those warrants count as a discount, so part of what OpenAI pays never shows up as revenue.
The upshot: The reported figure understates the true size of the OpenAI business, and that gap widens as the deal ramps.
2026-06-27 22:02:07
Welcome to the Saturday PRO edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
📊 Monthly reports: 200+ companies visualized.
📩 Tuesday articles: Exclusive deep dives and insights.
📚 Access to our archive: Hundreds of business breakdowns.
📩 Saturday PRO reports: Timely insights on the latest earnings.
Today at a glance:
🌐 Accenture: AI And Iran Hit Demand
🚚 FedEx: Premium B2B Pays Off
🛳️ Carnival: Europe Yields Soften
🫒 Darden: LongHorn Carries The Year
Accenture's Q3 revenue rose 6% Y/Y (3% in local currency) to $18.7 billion ($50 million miss), with GAAP EPS rising 9% Y/Y to $3.80 ($0.11 beat).
New bookings of $19.3 billion fell 2% Y/Y, the first year-over-year bookings decline since Q3 FY25. Shares plunged 18% post-earnings, the worst one-day drop on record, extending the stock's roughly 50% YTD decline.
CEO Julie Sweet flagged two distinct headwinds:
Middle East impact: $100 million Q3 revenue hit plus ~$400 million in sales impact as the Iran conflict slowed decision-making across EMEA. Sweet expects “more impact” in Q4.
Managed services deal slippage: A couple of large opportunities pushed into FY27 for company-specific reasons, creating timing-driven softness.
Accenture announced a $4.2 billion cybersecurity acquisition package: a majority stake in Dragos plus 100% of runZero and NetRise. Sweet called the combination "a first-of-its-kind OT Security platform." The three businesses generate ~$208 million in ARR with 48–53% Y/Y growth. Accenture also launched Accenture Edge, a new mid-market business targeting what it sizes as a $240 billion addressable market, and signaled it will tap the long-term debt market to fund elevated M&A. Total FY26 capital return was raised to at least $9.5 billion. Demand for “large-scale reinvention” continued: 104 quarterly bookings over $100 million year-to-date, up 13%.
Accenture cut the top end of FY26 revenue growth guidance to 3–4% (from 3–5%) and narrowed adjusted EPS to $13.78–$13.90 (from $13.65–$13.90). Q4 revenue forecast of $17.8–$18.4 billion fell short of the ~$18.5 billion consensus. CFO Angie Park said "more of the guided range" is in play, given macro uncertainty.
The stock now trades at its lowest multiple ever, roughly 6x EV/free cash flow. The combined dividend and buyback yield is near 10%. The big question is whether the AI displacement thesis continues to show up in slowing bookings, or whether the capital return profile is starting to make the bear case harder to justify at current levels.
2026-06-26 20:01:54
Welcome to the Free edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
For decades, memory was the invisible commodity inside every device: cheap, abundant, and easy to ignore. Not anymore. A brutal memory crunch is quietly pushing up the price of nearly every device you own, as the world’s top tech giants fight for supply and leave smartphones, PCs, and automotive lines competing for what remains.
That squeeze has a winner on the other side.
Micron just reported the most extraordinary quarter in its history.
Here’s what stood out.
Today at a glance:
☁️ Micron’s blowout quarter
🔒 Trying to break the cycle
🌐 The squeeze everyone else feels
🔭 What to watch next
FROM OUR PARTNERS
AWS built an infrastructure for the cloud and became one of the most valuable businesses in history. The segment made $128 billion in sales for 2025.
Now, another company is building what could be an equivalent infrastructure: a new highway for AI.
Traditional AI data centers are massive, take 3 to 7 years to build, and disrupt communities. But BluSky AI has designed a prefabricated SkyMod system: small footprints, powerful, and purpose-built for AI workloads. These data centers take months to stand up, not years.
Hundreds of thousands of enterprises use AWS to power their business. BluSky AI’s robust portfolio of sites will create new onramps for AI’s urgent needs. Invest in BluSky AI as they unlock a power source for 417M+ companies using AI.
This is a paid advertisement for BluSky AI Regulation A offering. Please read the offering circular at https://invest.bluskyaidatacenters.com/
Micron’s fiscal Q3 (ended May 28) was extraordinary by any standard.
Revenue: The top line surged +346% Y/Y to $41.5 billion ($5.6 billion beat).
☁️ Cloud Memory: $13.8 billion (+307% Y/Y).
🖥️ Core Data Center: $11.5 billion (+653% Y/Y).
📱 Mobile & Client: $11.5 billion (+254% Y/Y).
🚗 Automotive & Embedded: $4.6 billion (+311% Y/Y).
Margin trends: Gross margin was 85% (more than doubled Y/Y). Operating margin was 80% (+57pp Y/Y). These NVIDIA-esque margins drove earnings per share up 106% sequentially to $25.11 ($4.25 beat).
Cash flow: Operating cash flow was $25.4 billion (+451% Y/Y). Adjusted free cash flow was $18.3 billion (vs. just $1.9 billion a year ago).
Balance sheet: Cash, marketable investments, and restricted cash reached $30.2 billion. Long-term debt was $5.1 billion (down from $14.0 billion a year ago).
Q4 FY26 guidance: Revenue ~$50.0 billion (~$6 billion above consensus). Gross margin ~86%. EPS ~$31 (vs ~$25.72 expected).
Almost none of the outperformance came from selling more chips. DRAM bit shipments rose only in the low single digits, yet DRAM prices jumped more than 60% in a single quarter, and NAND prices rose in the mid-80s%. Micron is already capacity-constrained, so the shortage isn’t showing up in units. It’s showing up in price.
Now step back and look at the scale.
Micron booked more revenue in three months ($41.5 billion) than in any full fiscal year through 2024, and its data center business alone cleared a $100 billion annualized run rate. A year ago, total quarterly revenue was $9.3 billion.
The market has noticed. Shares are up more than 8X over the past 12 months and jumped as much as 16% after this print. A company worth roughly $150 billion a year ago is now worth close to $1.3 trillion. It’s the kind of move investors usually associate with platform companies, not memory cycles.
The numbers are staggering. The strategy behind them is the actual story.
Memory has always been boom-bust: demand surges, the survivors overbuild, prices crater, everyone bleeds. The cycles culled the field to three main players: Micron, Samsung, and SK Hynix. Their shared discipline on supply is what keeps it tight. For 40 years, the industry’s curse was that every boom financed the next bust. This quarter, Micron tried to turn customer panic into a new floor.
The mechanism is the Strategic Customer Agreement (SCA). It’s a multi-year, take-or-pay contract locking in both volume and price, unlike the one-year handshakes the industry has always run on. Micron has now signed 16, typically five-year deals running through 2030. They already cover ~20% of its DRAM and a third of its NAND, with management targeting 40%+ of total revenue. Customers are backing them with ~$22 billion in deposits and guarantees, including roughly $18 billion of it in cash.
The largest deals set a ceiling at today’s prices, and a floor beneath them. Micron says that even at the floor, gross margins would sit well above its prior cyclical peak of ~60%. For the covered volumes, the old ceiling is effectively the new floor. The minimum value of these contracts is ~$100 billion, which management calls a conservative number, not a forecast.
The most telling example is Anthropic, with a memory and storage supply agreement paired with Micron’s investment in the AI lab’s Series H round. A memory maker buying equity in the very demand it’s trying to lock down.
The pricing power makes those floors believable. Micron’s Mobile and Client unit — the part most exposed to squeezed phone and PC makers — ran an 87% gross margin, up from 24% a year ago.
Business model: Take-or-pay volumes and price floors smooth the industry’s most violent variable. The balance sheet already reflects the boom, and Micron now holds $24.4 billion in net cash. All three agencies upgraded Micron’s credit rating to BBB+.
Competitive moat: The counterparties are the hyperscalers and NVIDIA — the same names that decide HBM allocation. Locking them in deepens an engagement already hard to displace.
Investor angle: The bear case is always about what happens when the cycle turns. A floor margin above the old peak is the most direct rebuttal Micron has ever offered. The caveat is that upfront customer deposits are not free money. They come with future supply obligations.
For decades, Micron's curse was that good times invited the bust. By getting customers to commit volume, price, and cash years out, it's betting it can keep the boom without the hangover.
Micron’s 85% gross margin is someone else’s cost line.
The same shortage minting records in Boise is rippling through the rest of tech. Apple is raising prices, and HP’s memory bill doubled. Micron CEO Sanjay Mehrotra said on the call that there’s “no line of sight” to when supply catches up with demand. He pushed the crunch past 2027, a sharp reversal from earlier guidance that had it easing by next year.
It won’t ease quickly, because adding new supply is genuinely hard. Greenfield fabs take years and run into construction lead times, scarce skilled labor, permitting, and power. Meanwhile, each new node yields fewer incremental bits, and HBM’s heavier wafer appetite eats into everything else. Data-center memory is on track to top half the industry’s total bit demand for the first time this year. Everything non-AI is the residual.
But the edge isn’t only a victim of the crunch. Agentic platforms like OpenClaw and NVIDIA’s NemoClaw are pushing more AI workloads onto devices themselves, where better economics, privacy, and latency all matter. Micron expects that shift, paired with pent-up upgrade demand, to lift the memory inside each phone and PC over time. Today’s squeezed buyers could become the next source of memory demand.
The competition is sprinting after the same wave. SK Hynix, which leads the HBM market, just filed for a ~$29 billion US listing to fund its expansion, and Jensen Huang confirmed NVIDIA will source HBM4 for its next-gen Vera Rubin platform from all three memory makers. The allocation fight among Micron, Samsung, and SK Hynix is far from settled.
The real long-run wildcard sits in China. CXMT and Yangtze Memory are scaling fast, and standard DRAM is fungible enough that OEMs can qualify different suppliers across regions, product tiers, and configurations. Chinese suppliers still trail at the high end, especially in HBM, so this is not the 2026 story. But volume is how that gap closes.
🪜 The margin ceiling: Management itself flagged “a meaningful moderation in the rate of price increases” behind the ~86% gross margin guide for Q4 FY26. The step-up shrank to roughly +1.4pp from the ~+6pp it guided a quarter ago. Some read that as early price softening in 2H26 as SK Hynix adds supply and PC and phone volumes keep contracting. The counter is the SCA floor, which Micron says still clears its old peak.
🔒 How far above the floor: The contract terms are now largely public. Five years, take-or-pay, ~$100 billion minimum. The open questions are concentration (four large customers carry most of it) and how far actual revenue runs above that floor.
🌐 China and policy: Watch whether US export controls on China’s semiconductor equipment access tighten or loosen, and whether CXMT lands design wins at Western OEMs for Asia-bound products.
🏗️ The cost of staying ahead CapEx will jump to ~$10 billion in Q4 (~$27 billion for full-year FY26) and will step up again in FY27, with operating expenses set to rise ~$1 billion as R&D expands. The test will be whether Micron can fund the build without surrendering the balance-sheet discipline it just earned.
Bottom Line: Micron is trying to answer the oldest objection to memory stocks: the boom always invites the bust. This time, customers are locking in supply years in advance because AI has turned memory into a strategic bottleneck. For consumers, that means higher device prices. For Micron, it means margins even NVIDIA would envy.
That's it for today.
Happy investing!
Thanks to Fiscal.ai for being our official data partner. Create your own charts and pull key metrics from 50,000+ companies directly on Fiscal.ai. Save 15% with this link.
Disclosure: I own AAPL, AMD, GOOG, and NVDA in App Economy Portfolio. I share my ratings (BUY, SELL, or HOLD) with App Economy Portfolio members.
Author's Note (Bertrand here 👋🏼): The views and opinions expressed in this newsletter are solely my own and should not be considered financial advice or any other organization's views.
2026-06-23 20:02:23
Welcome to the Premium edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
Three companies announced roughly $86 billion in acquisitions in a single week. None of these deals is really about buying revenue.
SpaceX, Salesforce, and Fox are buying control points (the layers customers touch every day) before the market structure hardens around them.
When technology shifts, those layers are the most valuable real estate there is:
The workflow: where work gets done.
The interface: where choices are made.
The distribution layer: where attention is monetized.
The data loop: that improves the product over time.
A scarce asset can justify a premium, but a rushed deal can torch billions. The whole difference comes down to one question: is the buyer acquiring a durable control point, or simply paying up for growth it could not build internally?
Are these companies buying the future, or buying time?
Let’s review.
Today at a glance:
🤖 SpaceX + Cursor
☁️ Salesforce + Fin
📺 Fox + Roku
SpaceX is acquiring Cursor parent Anysphere in a $60 billion all-stock deal, struck four days after its Nasdaq debut.
The headline number is enormous. Cursor was valued at $29 billion in November 2025, then explored funding above a $50 billion valuation in April 2026, before SpaceX exercised its right to buy the company outright at $60 billion.
But the structure matters: SpaceX is paying with stock, not cash.
2026-06-19 21:02:45
Welcome to the Free edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights
In case you missed it:
SpaceX just pulled off the largest IPO in history. The rocket maker was priced at a fixed $135 per share on June 12 and surged more than 50% in its first three days of trading, briefly touching an astounding $2.6 trillion valuation.
More are on the way. Anthropic and OpenAI filed confidentially earlier this month. The most valuable private companies on earth are racing toward public markets.
The temptation to invest is obvious.
These are the companies investors have been waiting years to own: rockets, AI models, frontier labs, category leaders. The kind of names that can make every other stock in your portfolio feel boring.
But there’s an old market joke that IPO stands for “It’s Probably Overpriced.”
The hotter the IPO, the more erratic the early trading can become.
Should you take the plunge?
Four decades of IPO history point to a boring answer: probably not yet.
Today at a glance:
📉 The IPO pop mirage
📊 What 40 years of data say
🧭 The five-rule playbook for IPOs
🚀 SpaceX, OpenAI, or Anthropic?
The IPO pop you read about in the headlines is almost never yours.
The offering price and the opening price are two different numbers. In most hot IPOs, individual investors pay the second one.
That gap is often celebrated as a successful IPO. But it also means the company could likely have sold shares at a higher price. The first-day upside went to whoever received an allocation, not to the business or most public investors.
In traditional IPOs, banks often underprice the deal, allocated investors capture the first-day pop, and the company effectively eats the difference. Most individual investors buy in the open market after the pop.
That difference can be extremely short-lived. Rivian priced its IPO at $78, closed its first day at $101, and kept climbing. A week later, it hit $172, up 70%, making it briefly the third-most-valuable automaker on the planet, ahead of Ford and GM. In retrospect, it was the peak. The stock is down more than 80% from that first-day close, near $16 today.
Cerebras priced its AI listing at $185 and closed its first day at $311, up 68%. Spectacular for whoever got the allocation. Much less helpful for anyone buying after the move, when the valuation already reflected the excitement. A month later? The stock is nearly 30% off its peak.
A monster pop is not necessarily a triumph. Sometimes it is simply a temporary mispricing, and the buyer at the open is on the wrong side of it.
SpaceX tried to make the process fairer. It skipped the usual book-building theater and reserved an unusual 30% of the deal for individual investors.
The stock rose more than 50% in its first three days of trading anyway. The valuation now reflects even more future success than the IPO price did, leaving less room for error for new investors.
Takeaway: The first-day spike is usually a transfer of wealth to people who are not you. The smaller the pop, the fairer the deal. But a fairer price and a good investment are not the same thing.
IPOs tend to underperform. And by a lot.
Why? IPO investing has three built-in disadvantages:
Limited operating history: The S-1 shows the past, not how the company behaves under public-market pressure.
Insider selling: Lock-ups often expire after ~6 months, when employees and early investors can finally sell.
Information edge: The best investors often saw the company years earlier, at much lower valuations, with better access.
Jay Ritter has tracked new listings for decades. The University of Florida finance professor built one of the definitive IPO datasets.
Across thousands of operating-company IPOs with at least $100 million in sales, measured from the first-day close, the five-year record is sobering:
50% posted negative returns over the five years after their first-day close.
30% lost half their value or more.
Only 24% became multi-baggers, meaning they gained 100% or more.
This data excludes the smallest IPOs, those with less than $100 million in trailing sales. Including those would make the record look even worse, but also less relevant to the kind of large, category-defining companies most investors care about.
Finishing above the first-day close is not the same as beating the market. A stock that rises 20% over five years may look fine in isolation, but it still failed as an allocation if the S&P 500 did much better over the same period.
That’s why IPO researchers also look at market-adjusted returns. On that basis, the IPO record is even less forgiving. Larger IPOs do much better than tiny speculative listings, but on average, they still underperform comparable public companies after the first-day pop.
The damage clusters in a specific window: IPOs lag comparable companies most from 6 to 24 months after listing, around the time the lock-up expires and insiders are finally allowed to sell.
As the novelty fades and the selling pressure lands, expectations come back to earth, and so does the price.
If we expand to a 3-year window, several patterns show up again and again in Ritter’s 40-year data set:
Size matters: IPOs with under $100M in sales trailed the market by ~34%, while those with over $500M in sales trailed by just ~4%.
Profitability helps: Profitable issuers underperformed by ~13%, while unprofitable ones underperformed by ~31%. None of the megacaps heading for IPOs are profitable.
VC backing boosts the odds: Venture-backed names trailed by ~14% — better than the ~25% for everyone else, but still behind.
Takeaway: A new listing is usually a stock priced for the long-term growth story, at the time of maximum execution risk. Time separates the winners from the wreckage. Buying a great business can be sound, but the IPO is often the worst moment to do it.
Warren Buffett has long been skeptical of IPOs. His view is simple: IPOs come to market when sellers choose the timing, not when buyers are likely to get the best deal. He once put it bluntly:
“It isn’t worth spending five seconds thinking about IPOs.”
Of course, an IPO is merely a moment in time, not an asset class. What looks uninvestable at the IPO price can eventually become a compelling idea.
That’s why having a clear process matters.
Here are a few rules I’ve created for myself:
Avoid the IPO hype: The first-day pop mostly belongs to investors who received the allocation, not to open-market buyers. Rushing in on Day 1 forces you to pay a hype-driven premium at the moment of maximum risk. The rule is simple: let the initial open-market volatility pass without touching it.
Wait for the second earnings call: The S-1 is a snapshot. Public markets are the test. The first earnings call shows how management talks to investors. The second shows whether they can forecast appropriately. Are the right KPIs improving? Is growth accelerating or decelerating? Did guidance hold? Did the lock-up create selling pressure? Two quarters help draw a trendline. Anything earlier is closer to buying blind.
Nibble in year one: The first year is usually the noisiest. The stock is absorbing new shareholders, lock-up expirations, analyst coverage, guidance resets, and the first real tests of public-market expectations. A small starter position lets you follow the company closely without making the IPO price your entire cost basis. If a 50% drawdown would damage your portfolio or your sleep, the position is too big.
Anchor to valuation: Missing a 20% move is not a disaster. Buying a great company at a price that already assumes perfection can be. The question is not whether the company is exceptional. It is whether the valuation leaves room for error. Don’t sweat daily movements; instead, focus on valuation relative to fundamentals.
Give it time: Many great stocks go nowhere in their first few years as public companies. That doesn’t mean the business is broken. It means the market is still figuring out the right multiple, the right expectations, and the right shareholder base. Multi-baggers rarely require buying on day one. They require surviving the early volatility.
Takeaway: The playbook is boring on purpose. Wait for public evidence, size small, and let the lock-up cycle pass. Boring is how you beat the IPO base rates.
Neither of the new-frontier IPOs has demonstrated how the business behaves under public-market pressure. Their margins at maturity are still uncertain, and their future growth is far from guaranteed.
None of this means the companies are bad. It means the setup is hard.
These may become some of the defining franchises of the next decade. The case for owning the leaders of AI and space is real. But the case for owning them on the first day, at peak-euphoria valuations, before a single public quarter, is much weaker.
The better move is boring: put them on your watch list.
Let them report. Let the first wave of excitement pass. Let insiders sell. Let expectations move from story to numbers. Then decide whether the business is getting stronger and whether the valuation leaves room for error.
Some of these stocks may compound for a decade. The playbook is not to ignore them entirely. It’s to invest on your own terms, at a price set with a clearer head.
Bottom Line: The 2026 IPO wave is a readout of how hot the market has become, not a buy signal. SpaceX, OpenAI, and Anthropic belong on the watch list, not the impulse-buy list. The best businesses give you years to get in. The IPO only sells you the urge to rush.
Once Anthropic and OpenAI’s S-1s are available, you can expect our breakdowns and signature visuals. Stay tuned for that!
That’s it for today!
Happy investing!
Thanks to Fiscal.ai for being our official data partner. Create your own charts and pull key metrics from 50,000+ companies directly on Fiscal.ai. Save 15% with this link.
Disclosure: I own ABNB, BABA, CPNG, DASH, META, SNOW, and UBER in App Economy Portfolio. I share my ratings (BUY, SELL, or HOLD) with App Economy Portfolio members.
Author's Note (Bertrand here 👋🏼): The views and opinions expressed in this newsletter are solely my own and should not be considered financial advice or any other organization's views.
2026-06-16 20:01:07
Welcome to the Premium edition of How They Make Money.
Over 300,000 subscribers turn to us for business and investment insights.
In case you missed it:
SpaceX just pulled off the largest IPO in history, raising about $75 billion at a valuation near $1.8 trillion. The stock has already surged more than 40% in its first two days, pushing its market cap past $2.5 trillion, nearly as much as Amazon.
But it was only the opening act.
Anthropic and OpenAI have already filed confidentially to go public, setting the stage for massive IPOs in the coming months. Both could list above $1 trillion.
Anthropic’s private valuation has rocketed past OpenAI’s to $965 billion, making the Claude maker, on paper, the most valuable standalone AI lab on earth.
But the company just got a reminder of how fragile that value can be. On Friday, Anthropic received a US export-control directive restricting access to its two most powerful models. Because it could not screen users in real time, the company disabled both models worldwide.
Every major AI company is choosing where to sit on the build-vs-rent spectrum:
Build a frontier model like Anthropic, OpenAI, and Google.
Rent one from a company that already has.
Until last week, that looked mostly like a trade-off between cost and control. Now there’s another variable: the risk of regulatory shutdown.
Today at a glance:
The frontier’s very bad week
Why building just got riskier
Apple’s accidental hedge